Invest3 publishers3 min readPublished
Software stocks rallied on Amodei's slowdown essay before Huang called it a false choice
Enterprise software soared on an essay whose first step Anthropic had already committed to. Nvidia's answer accepts independent auditors and rejects new statutes, and the cash difference lands in accelerator demand.
The Investor · Invest desk
What happened
- Amodei published an essay on the 12th arguing for slowing the pace of AI development, then defended it on Benioff's Dreamforce keynote stage in San Francisco on the 15th, in front of about 12,000 people.
- Huang followed him on the same stage and said market forces already exist and the industry does not need new laws or regulations, describing safety as an engineering problem.
- OpenAI's Sam Altman, SpaceX's Elon Musk and Google DeepMind chair Demis Hassabis quickly backed Amodei's plan. Amodei framed it as tempering capability gains without sacrificing commercial advantage or the US lead.
- At the All-In Summit around the same time, President Trump joined Huang by phone, called AI safety concerns a hoax and turned to the economic importance of data center expansion, according to Cryptobriefing.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint Amodei's first step is a commitment Anthropic has already made and the remainder is a conversation with the industry, so nothing in the plan obliges a rival to shorten a training run or cancel an order.
- exposure Nvidia's hardware demand tracks the pace of development, so only a slowdown that changes training budgets touches its revenue; an essay and a stage disagreement do not.
- contradiction Huang rejects new statutes but accepts independent evaluators on something like an audit model, which is close to the industry standard-setting Amodei describes, so the split is narrower than global rules against no rules.
- decision Anyone who bought enterprise software on Monday's move is holding a position that requires model progress to actually decelerate, and has to re-mark it if the slowdown stays voluntary.
Monday's bid was for the application layer. Enterprise software stocks soared on Amodei's call to slow AI development [14]. The group was crushed earlier this year on AI disruption fears and has been recovering as Wall Street warms to the idea that AI cannot replace everything [18]. Salesforce is up more than 50% from its 52-week low of $146 in late June [16]. That puts the shares above roughly $219 [17]. On Tuesday it handed back part of a more than 4.5% advance [15].
The document those buyers priced asks for less than the tape implies. "Step one is what we have already committed to, and for the rest we plan to talk with the industry," Amodei said on stage [4]. The cross-border piece comes last. "Finally, an international element has to be included," he said, calling that the core of his three-step plan [5].
Huang was blunter. "It's a false choice," he told Benioff. "You could definitely have both at the same time" [7]. His version of pacing is internal and voluntary. "Run as fast as you can," he said, "But if you feel at any given point in time the company's out of control or the product's not going to be safe, take a pause and make sure you get it right" [9]. According to Cryptobriefing, he favours independent evaluators and testing regimes closer to financial audits than federal mandates [11], and he called predictions of human extinction risk from AI "made up" [12].
The cash difference sits with Nvidia. Cryptobriefing notes the company is the dominant supplier of AI accelerators, and that demand for its hardware rises when AI development accelerates and slows when development slows [13]. The same publication argues the weak point in Huang's framework is who decides when to pause. By that argument it depends on companies recognising a safety problem and stopping on their own, and market incentives work against that [32]. Salesforce is buying from both positions inside a month: a new Agentforce reasoning model built on one of Nvidia's open-weight Nemotron models on Tuesday [19], and Claudeforce with Anthropic late last month [20]. Benioff, asked by reporters how model companies should proceed, said "If they feel like they should slow down, then they should slow down," and added that "they should be held accountable" [22].
I think Monday's move was a price paid for a document, not for a change in anyone's spending. The counter-case is serious. An essay endorsed within days by the people running the other frontier labs is how an industry norm starts, and a genuine tempering of frontier training runs would reach accelerator demand before it reached anyone's ethics statement [13]. Two things would show me wrong. A lab guiding capital spending down, or a standards commitment with a signature and a named counterparty attached to steps two and three [4].
What to watch
- A frontier lab guiding capital spending or training budgets down. That is the first place a real slowdown would be visible.